VAST DataEnclave brings end-to-end protection for enterprise AI workloads
“What we're building is to try to close this gap so that we can have an end-to-end secure confidential environment from data at rest to data in transit and data in use,” says John Mao, VP of Business Development, VAST Data.
As organizations look towards developing more AI use cases, having control over their data is imperative. This not just includes having visibility over data but ensuring that sensitive data remains on-premise or in environments that can be controlled by the organizations.
In a media briefing, John Mao, VP of Business Development, VAST Data explained that while AI will be fairly accessible to everybody, what makes AI different is when it's applied into an enterprise to their own proprietary data.
“It is the most safeguarded thing. In regulated industries, there's challenges, and there's government regulations to make sure that information doesn't leave the country, or sometimes it's heavily restricted for a variety of reasons. At the same time, those enterprises want access to AI. And so, there's this dilemma that they have today, where a lot of enterprises just don't have a lot of choice, because leading AI is somewhere else, it's usually hosted in the cloud, it's somewhere as a SaaS offering, and so forth,” Mao explained.
Mao added that enterprises that want to store data somewhere, the service provider needs to be able to guarantee that they can't see your data which is data at rest.
For data at transit, Mao explained that when an organization needs to move data from a storage system into a processor or into a computer machine, they need to ensure that nobody can inspect the network and steal the data in transit.
While VAST Data has capabilities to solve both data at rest and data at transit concerns, Mao pointed out that the missing piece that VAST didn't have was data in use or data while it's being processed.
“So if you have the data stored, it's secured while the data is in transit and it's secured once the data gets into the machine, whether it's a classic server or a GPU accelerated server, as soon as that data is being processed and analyzed, that's an open risk area where if a rogue actor could get in, they could actually still see the data because it's being processed in the open,” he said.
VAST DataEnclave
To help organizations with this, VAST Data has collaborated with NVIDIA and its partners to bring proprietary and open AI models together across a range of modalities with customers having control over cost, performance, model selection and data privacy.
“What we're building is to try to close this gap so that we can have an end-to-end secure confidential environment from data at rest to data in transit and data in use,” Mao said.
The VAST Data Enclave, which is the confidential AI capability of the VAST DataEngine, enables deployment inside customer data centres or trusted cloud hardware, including environments where leading AI models could not previously operate.
VAST DataEnclave ensures customer data keys remain under customer control, model keys and weights remain within the model builder’s trust domain, and infrastructure operators and administrators cannot access either while they are being processed.
Put simply, VAST DataEnclave enables AI cloud providers worldwide to deliver attested, sovereign environments where model builders, enterprises, and governments retain control over their IP and data within their own jurisdictions. It relies on NVIDIA Confidential Computing to create a secure container runtime and attestation service directly within the VAST DataEngine. Proprietary models execute inside secure enclaves established through CPU and GPU trusted execution environments.
“We believe that there is a massive unlock from a market opportunity perspective for a couple of different use cases. VAST DataEnclave allows model builders to now bring their solution to an enterprise customer wherever they want to run, whether it's in some cases, it might be even on-prem. In other cases, more likely it would be within a cloud provider and a partner to be able to run that there. And so, both sides win at the end of the day when it comes to enabling this kind of uniform trust model or zero trust model across the entire ecosystem,” Mao said.
For James Manning, CEO and Co-founder, Sharon AI, a leading Australian neocloud, delivering trusted, sovereign AI infrastructure to organizations that need to build, train and run AI at scale, Vast DataEnclave builds on their sovereign data foundation with VAST, enabling them to host those models onshore, inside attested environments where the model owner’s weights and the customer’s data are both protected from everyone.
“What I will say is the product enables, from our perspective, us to pursue some of that software compute aspect. And it's a very important part of our growth profile as we think about our business and the partners we pursue, the customers we pursue with our partners,” Manning said.
Meanwhile, Sunil Chavan, VP for APAC at VAST Data shared that VAST has had conversations with potential customers in Singapore that are looking to work with partners like Sharon AI to see how they can start using Vast DataEnclave, understand the use cases and have a clear focus on secure, integrated, model-trained data.